AI is burying the IRS in comments. Yours just got harder to hear.
When Treasury proposes a tax regulation, the public gets to talk back. That comment file is where practitioners flag the unworkable definition, the missed edge case, the compliance cost nobody in Washington modeled. It is one of the few places where a well-argued letter from a working CPA can actually change the rule. This week the Government Accountability Office said, in effect, that the channel is drowning.
In a report published Tuesday, the GAO found that Treasury and the IRS risk being overwhelmed by the volume of public comments on proposed tax regulations, sometimes tens of thousands; a single proposed digital-asset regulation drew 44,821. The agencies, the GAO concluded, are not following leading practices for public engagement in rulemaking. The AI finding is the part to sit up for: members of the public reported using AI to help write comments, and, in the GAO’s words, “AI can write slightly different comments on the same issue for each person, making near-duplicate comments harder to identify.” The IRS, per the report, faces increasing challenges identifying duplicates at all, and has no policies for handling mass comments or AI-written ones.
The timing is not incidental. As Accounting Today noted, the wave follows last year’s One Big Beautiful Bill Act, which set off a run of proposed regulations on deductions for tips, overtime, seniors, and car loans, exactly the kind of kitchen-table provisions that generate mass campaigns. Treasury’s response came from Kenneth Kies, the acting IRS chief counsel and assistant secretary for tax policy: the agencies “appreciate GAO’s review and remain committed to complying with all applicable rulemaking requirements.”
The GAO’s fix is procedural: six recommendations, three of them aimed at the comment problem, including policies for spotting identical and nearly identical comments, AI-generated ones included, and better virtual participation in public hearings. The detail that deserves more attention sits in the appendix. In its written reply, Treasury agreed that a robust process to consider public comments is crucial to rulemaking, then disagreed with the recommendation to build those mass-comment policies, and with both hearing recommendations. The agency the GAO describes as at risk of drowning has, for now, declined the life raft. And none of it changes the underlying economics: generating ten thousand plausible, slightly varied comment letters now costs almost nothing, and reading them still costs the government real analyst hours.
Here is what that means if you are the one commenting. The value of a comment was never its word count; it was always the technical catch, the worked example, the client fact pattern the drafters had not seen. That kind of letter is now more valuable, not less, because it is the signal in a growing pile of synthetic noise. But it only wins if it is visibly the real thing. Put your name, your credential, and your firm on it. Lead with the specific provision and the specific failure. Attach the numbers from an actual engagement, suitably anonymized. A letter like that cannot be mass-produced, and it is the one an overwhelmed reviewer can least afford to skip.
There is a wider pattern here, and regular readers will recognize it. On Tuesday we wrote about the audit regulator rewriting its five-year plan around AI while its own members argue about the technology. Two days later, the GAO described the tax side of the same problem: institutions built for human-scale input, meeting machine-scale volume, with no policy in the drawer. The tools arrived faster than the process did.
The bottom line: AI just made it free to sound like ten thousand people. It cannot make anyone accountable for what those people said. A comment letter is a professional putting their name behind an argument, and that is the part no model can generate. The noise is about to get worse. Make sure your letter is unmistakably signal.
— Footnote
Footnote is an independent publication. It is not professional accounting, tax, or legal advice. This piece is based on the GAO report and public reporting linked above; agency statements are their own. Details are current as of July 24, 2026.